On a weighted mean square result of the error terms of some arithmetical functions (Q1969345): Difference between revisions

Let \(F(s) = \sum_{n=1}^\infty a_nn^{-s} (s = \sigma + it) \) have a meromorphic continuation for \(\sigma > \sigma_0 - \varepsilon_0\) for some \(\sigma_0 \geq 0\) and \(\varepsilon_0 > 0\). The author's main result is the asymptotic formula \[ \int_1^T|E(y)|^2y^{-2\sigma_0-1} \text{ d}y \sim c\log T \qquad(c > 0, T\to\infty),\tag{1} \] where \(E(x)\) is the error term in the asymptotic formula for the summatory function of \(a_n\), and \(F(s)\) satisfies some fairly general conditions. To prove (1) the author expresses first \(E(x)\) by Perron's formula as an integral over the line \(\Re s = \sigma_0 + \delta (\delta\to 0^+)\), sets \(x = \text{ e}^u\) and then uses Parseval's formula for Fourier transforms to express \[ \int_{-\infty}^\infty |E(\text{ e}^u)|^2 \text{ e}^{-2(\sigma_0+\delta)} \text{ d}u\qquad(\delta\to 0^+)\tag{2} \] in terms of a mean square integral involving \(F(s)\). If \(F(s + \sigma_0) = \zeta(s)G(s)\), where \(G(s)\) satisfies some hypotheses, one obtains (1) by applying a standard Tauberian theorem, after suitable transformations of the expression for (2). Several interesting applications of (1) are given, including \[ \int_1^T E^2_a(y)y^{-1-2(a+|a|)} \text{ d}y \sim C_a\log T \qquad(C_a > 0, T\to\infty) \] with explicit \(C_a\), where \(a_n = \sigma_a^2(n)\) for \({1\over 2} < |a|< 1\), and \(\sigma_a(n) := \sum_{d|n}d^a\).